Anthropic's $1.5 Billion Book Piracy Settlement Sets Precedent for AI Training Data Acquisition

In a landmark ruling, a federal judge in San Francisco approved a $1.5 billion class-action settlement between AI company Anthropic and authors and publishers whose books were downloaded from pirate sites like Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi) to train its Claude chatbot. The July 20, 2026, final approval by U.S. District Judge Araceli Martínez-Olguín marks the largest known copyright recovery in U.S. history, covering approximately 482,000 to 500,000 works at roughly $3,000 per title before fees and expenses.[1][2]

The case, Bartz v. Anthropic, filed in 2024 by authors including Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, hinged on a key distinction drawn by now-retired Judge William Alsup. In a June 2025 ruling, Alsup held that training AI models on copyrighted books constitutes fair use under Section 107 when the copies are lawfully acquired, describing the purpose as "exceedingly transformative." However, he ruled that Anthropic's creation of a permanent "central library" containing over seven million pirated books was "inherently, irredeemably infringing," exposing the company to potential statutory damages up to $150,000 per work.[3][4]

Anthropic settled in September 2025 to avoid trial, agreeing to pay the settlement fund and destroy the pirated datasets. Over 91% of eligible rightsholders have claimed their shares. The settlement releases claims related to past acquisition but leaves open questions about future outputs or new lawsuits, including separate actions by opt-out plaintiffs and music publishers.[5][6]

This outcome underscores tensions in AI development: while transformative use for analysis receives protection, the method of data acquisition does not. It provides a concrete benchmark for licensing costs—far exceeding typical per-work infringement awards—and highlights diligence failures in sourcing training corpora. Related suits against Meta, OpenAI, and others continue, with the Copyright Office's 2025 report echoing concerns over market substitution. The settlement avoids setting broad precedent on fair use but signals that "shadow library" sourcing carries significant financial risk.[7]